Kepler emerges with $468M to tackle the AI memory bottleneck
Kepler Computing emerged from stealth with $468 million raised and a new approach to high-bandwidth memory for AI systems.
The AI chip shortage is increasingly becoming a memory problem as well as a GPU problem.
What happened
Kepler Computing emerged from stealth after raising $468 million over several years from investors including Intel Capital, AMD Ventures and Gates Frontier Fund. It is developing memory architectures using ferroelectric materials and 3D stacking, with production plans involving existing semiconductor fabs.
Why it matters
High-bandwidth memory is one of the most constrained components in modern AI systems. A technology that improves bandwidth and energy efficiency without requiring entirely new fabs could become strategically valuable.
The bigger picture
AI infrastructure investment is spreading into every part of the compute stack. Startups targeting memory, packaging and interconnects are attracting increasingly large pools of capital alongside accelerator companies.